The synthesis of emotions in artificial intelligences: an affective agent architecture for intuitive reasoning in artificial intelligences

The synthesis of emotions in artificial intelligences: an affective agent architecture for intuitive reasoning in artificial intelligences

[Abstract]: This dissertation addresses several highly-critical issues in affective computing and agent architecture design including knowledge representation, motivation, emotion appraisal and affective decision making. The approach presented integrates motivational drives, goals and associated behaviours via a multi-dimensional Affective Space. The research focuses on an emotionally motivated artificial intelligence (EMAI) architecture. This architecture dispenses with the ideas implemented in contemporary affective agent architectures where individual emotional states are modelled as individual variables, integrated and processed using complex algorithms. Contemporary approaches required significant programming effort to modify them for domains outside their realm, integration of new emotional states and high-level complex affective decision making. Unlike contemporary affective agent architectures, the EMAI architecture reasons using a multi-dimensional decision making process where emotional states are modelled as coexisting locations in a six-dimensional affective continuum called the Affective Space. Through use of the Affective Space, an EMAI agent can predict the effect that certain behaviours will have on its emotional state and in turn decide how to behave. Furthermore, the agent can use the emotions produced from its behaviour to update its beliefs about particular events and entities. The nature of the Affective Space also allows an EMAI agent to deal with processes related to emotion synthesis in a more effective manner than contemporary architectures. These processes include the natural diminishing of an emotional state's strength over time, the way in which emotions can influence an agent's perspective of a situation and the way in which an agent can migrate from one emotional state to another. This dissertation contributes crucial and unique concepts and formalisations of emotion based intelligence for agent construction to the domain of Artificial Intelligence (in particular Affective Computing). It introduces a unique process for emotionally motivated decision making based on holistic and atomic appraisals made with respect to events. The thesis contained within has been supported through experimentation that has confirmed the effectiveness of the emotion synthesis technique in the EMAI architecture and how this is used to produce intelligent agents capable of emotional reasoning and decision making.